Eligible rows before the holdout.
MODEL 3 · STEP 4Train
Train
through time.
Select among three compact ridge families on development folds only, then audit 252 terminal labels spanning 272 sessions under the corrected protocol.
MARKET CONTEXT · OUTSIDE MODEL FREEZENKE
These readings update independently. They are not model inputs, deployment evidence, or a trading recommendation.
- Latest price
- $38.40 · fallback
- 20D momentum
- -7.9% Close-to-close · 20 completed sessions
- RSI 14
- 40.1 neutral · Wilder method
- Relative volume
- 0.7× Versus the prior 20 completed sessions
TECHNICALS THROUGH · TIMESTAMPED FALLBACK
Choose on development evidence; audit on terminal labels.
Three compact L2-logistic families are ranked by pooled development out-of-sample Brier score. 10 expanding outer folds use a five-session purge; the final 252 labels span 272 market sessions and are excluded from candidate selection, tuning, scaling, and calibration.
Predictions used for family selection.
2025-06-24 to 2026-07-23 · 272 sessions.
Expanding, purged chronology.
Block-bootstrap training runs with declared seeds.
Point-in-time fundamental ridge.
| Candidate | Brier | Improvement | AUC | ECE | Probability std. |
|---|---|---|---|---|---|
| Point-in-time fundamental ridge · selected | 0.253137 | +0.000718 | 0.556 | 7.5% | 4.0% |
| Cross-market ridge | 0.253623 | +0.000233 | 0.531 | 6.7% | 2.9% |
| Compact price ridge | 0.254112 | -0.000257 | 0.528 | 6.9% | 2.7% |
The earlier provisional run exposed timing and target-policy defects. This corrected protocol fixes them before rerunning, but the same terminal period was already inspected. Treat the result as a reproducible audit-corrected run, not fresh independent confirmation.
NEXTStep 5 · Stress the result
Continue through the audit-corrected Model 3 workflow.